JamesHWade / shinymcp

What is shinymcp?

shinymcp lets R Shiny developers convert their interactive apps into MCP Apps—UIs that embed directly in AI chats like Claude Desktop. Instead of rewriting dashboards for LLMs, you flatten Shiny's reactive graphs into MCP tools that update outputs on input changes, with components mirroring selectInput or plotOutput. Run `convert_app("path/to/shiny")` for auto-scaffolded code, or build manually with `mcp_app(ui, tools)` and `serve(app)` over stdio or HTTP.

Why is it gaining traction?

It bridges the massive Shiny ecosystem to the MCP protocol, letting data apps like Palmer Penguins explorers live inside AI without JavaScript rewrites or page reloads. The parse-analyze-generate pipeline handles most conversions automatically, spitting out editable tools with placeholders, while a bundled AI skill tackles edge cases like modules. Zero npm deps and bslib theming make it drop-in for R teams eyeing agentic workflows.

Who should use this?

Shiny devs building dashboards for data analysis or stats viz, especially those integrating with Claude Desktop for LLM-driven exploration. R data scientists wanting to expose mtcars/iris summaries or ggplot scatters to AI agents without standalone hosting. Teams prototyping AI tools where Shiny logic becomes callable functions.

Verdict

Grab it if you're in the R Shiny + AI niche—solid docs, examples, and auto-conversion make early experiments viable despite 19 stars and 1.0% credibility signaling experimental status. Polish tool bodies manually for production; test coverage is decent but adoption will dictate longevity.

(198 words)